[ont.events] ICR Colloquium

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (06/30/89)

                 The University of Waterloo
                   200 University Avenue
                     Waterloo, Ontario


         The Institute of Computer Research (ICR)

                  Presents a Colloquium on



                        Organized C


by  Mr. Jiri Soukup

of  Code Farms, Inc., Richmond, Ontario




ABSTRACT

Organized C is a simple addition to C (or C ++) which allows automatic
management of data structures.  From a theoretical point of view, Organized C
is a parametric type library of organizations with a convenient and efficient
interface.  It  forms the organizations in top-down fashion, as compared to
object-oriented languages based on classes that work essentially bottom-up.
Organized C is a spinoff from VLSI CAD, but is applicable to any general C
program.  It is useful for complicated algorithms and for rapid development of
in-core databases which rely on a network of pointers.

After the presentation, there will be a demonstration for those interested.



Wednesday, July 5, 1989
3:30 p.m.
William G. Davis Computer Research Centre, Room 1302

Everyone is welcome.  Refreshments served.

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (07/26/89)

                 The University of Waterloo
                   200 University Avenue
                     Waterloo, Ontario


         The Institute of Computer Research (ICR)

                  Presents a Colloquium on


   A Unified View of Propositional Knowledge Base Updates



by   Hirofumi Katsuno

of   NTT Basic Research Laboratories, Tokyo, Japan




ABSTRACT

The semantics of revising knowledge bases represented by sets of
propositional sentences is analyzed from a model-theoretic point of
view.  The operation which incorporates new knowledge into knowledge
base is called  revision.  A characterization of all revision
schemes that satisfy the Gardenfors rationality postulates is
given in terms of an ordering among interpretations.  A different kind
of change occurs when a sentence previously believed becomes
questionable; the operation that makes this change is called
contraction.  Properties of the contraction operator that can be
defined in terms of revision are also studied.  Two new update
operators, elimination and recovery, are introduced.
Elimination discards all previous preconceptions on a set of
propositional letters; recovery undoes the effect of the last update.
It is shown that elimination cannot be expressed as a contraction, and
that recovery is in general impossible.  The existence of an invariant
part of the knowledge base comprising a set of integrity constraints
is considered and the definition of revision and contraction are
modified to take integrity constraints into account.

This is joint work with Alberto O. Mendelzon.



Wednesday, August 2, 1989
3:30 p.m.
William G. Davis Computer Research Centre, Room 1302

Everyone is welcome.  Refreshments served.

ksbooth@watcgl.waterloo.edu (Kelly Booth) (09/07/89)

Massively Parallel Processing for Graphics

Dr. Frank Crow
Principal Scientist Palo Alto Research Center

Date:  Wednesday, September 13, 1989
Time:  3:30 pm DC 1302
Place: Davis Centre, Room 1302

Abstract

Many architectures and algorithms have been proposed for applying
massively parallel methods to computer graphics.  In recent years the
onrush of technology has stampeded those of us who like to think about
such things from the realms of fantasy and wishful thinking to the
realm of the actually possible.  A few massively parallel graphics
systems have now actually been implemented.  We can expect to see many
more very soon.

Can we really expect to attain, through parallel systems, the six
orders of magnitude speedup necessary to produce today's most expensive
imagery in real time?  Answers may lie in looking at some current
approaches to massive parallelism in graphics and the bottlenecks they
leave.  It will also help to look carefully at what is required to make
images and how information must flow from shape descriptions to
pixels.

Refreshments

The audience is invited to attend a wine & cheese reception at 4:30 pm
in the ICR Lounge (DC 1301) immediately after the colloquium.  This
will be in lieu of the customary coffee and squares prior to the
colloquium.  A selection of wine, beer, and non-alcoholic beverages
will be available, accompanied by fruit and cheese.  The reception is
hosted by the Institute for Computer Research, the Department of
Computer Science, and the Computer Graphics Laboratory.

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (09/18/89)

ICR presents a colloquium on The Parallel Solution of Initial Value Problems
For Ordinary Differential Equations, by Dr. Kenneth R. Jackson of the
Computer Science Department, University of Toronto.  Davis Centre, Room 1302
Wednesday Sept. 20, 1989 at 3:30 p.m.  Usual refreshments will be served.
.
.

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (10/06/89)

ICR presents a colloquium on Abstraction in Artificial Intelligence Planning
with Dr. Qiang Yang, Department of Computer Science, UW.  Wednesday
October 11, 1989 at 3:30.  Davis Centre Room 1302 - refreshments will
be served.  Everyone welcome.

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (10/18/89)

ICR Colloquium

Free-form Modeling with Implicit Quadratic Surfaces

Wednesday, October 18, 1989 at 3:30 p.m.
Davis Centre Room 1302

Dr. Joe Warren
Department of Computer Science
Rice University, Houston, Texas

Everyone welcome.  Refreshments will be served.

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (10/19/89)

ICR Colloquium

A Practical Theory of Programming

Dr. Eric C.R. Hehner
Department of Computer Science
University of Toronto

Date: Wednesday, October 25, 1989
Time: 3:30 p.m.
Place: Davis Centre Room 1302

ABSTRACT

The logic of programs that we present is both simpler and more general than its
competitors.  In place of a pair of predicates (precondition and post-condition)
as in Hoare Logic, or a function from predicates to predicates as in Dijkstra's
predicate transformers, we offer a single predicate to serve as specification
and as semantics.  We do note require a Kleene squence or a least-fixed-point
construction or any induction that is specific to programs or computation.  We
include time complexity.  The logic covers sequential and parallel programming
communcating processes and nonterminating computations.

Everyone welcome.  Refreshments served.

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (10/27/89)

ICR Colloquium

HIgh Level Synthesis of Digital Hardware

Dr. Raul Camposano
IBM Thomas J. Watson Research Centre

Date: Wednesday, November 1, 1989
Time: 3:30 p.m.
Place: Davis Centre, Room 1302

Abstract

High level synthesis is the automatic design of a register-transfer
level structure that realizes a formally specified behaviour.  This lecture
gives an overview of the different issues involved in high-level synthesis
, i.e., design representation, high-level optimizations, scheduling,
allocation and interface to other design tools, mainly logic synthesis.
It emphasizes the two central problems in high-level synthesis: scheduling
and allocation.  In synchronous designs, scheduling consists of assigning
operations to discrete time slots called control steps.  The hardware for the
execution of the operations is defined during allocation.  Scheduling
and allocation are interrelated and depend on each other.  The main
algorithms for these transformations developed in Yorktown are shown.  Finally
some open problems such as design verification, design with constraints and
pipeline synthesis are addressed.

Coffee and cookies will be served.  Everyone welcome.

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (11/04/89)

ICR Colloquium

Inductive Reasoning and Kolmogorov Complexity

Dr. Paul M.B. Vitanyi

Centrum voor Wiskunde en Informatica
The Netherlands
&
Universiteit van Amsterdam
Faculteit Wiskunde en Informatica

Date: Wednesday, November 8, 1989
3:30 p.m.
Davis Centre, Room 1302

Abstract

Reasoning to obtain the `truth about reality, from external
data, is an important, controversial and complicated issue in
man's effort to understand nature.  Yet, today, we try to make
machines do this.  There have been old useful principles, new
exciting models and intricate theories scattered in vastly
different areas including philosophy of science, statistics,
computer science and psychology.  We focus on inductive
reasoning in correspondence with the ideas of Solomonoff.
While his proposal results in perfect procedures, they involve
the noncomputable notion of Kolmogorov complexity.  We develop
the thesis that Solomonoff's method is fundamental in the
sense that many other inductive principles can be viewed in 
particular ways to obtain comutable approximations of the 
method.  We demonstrate this explicitly in the cases of
Gold's paradigm for inductive inference, Valiant's learning
(by adding computational requirements), Rissanen's principle
and Jaynes' maximum entropy principle.  We present several
new theorems and derivations to this effect.  We also delimit
what can be learned and what cannot be learned in terms of
Kolmogorov complexity and we describe an experiment, in
machine learning of Kolmogorov complexity and its applications,
now in progress.

machine 
exerep

ylkingsbury@watdragon.waterloo.edu (Yvonne Kingsbury) (11/17/89)

                 The University of Waterloo
                   200 University Avenue
                     Waterloo, Ontario


         The Institute of Computer Research (ICR)

                  Presents a Colloquium on


    Using Spatial Coherence as a Local Teacher for a Neural Network


by    Dr. Geoffrey Hinton

of    Dept. of Computer Science and Psychology, University of Toronto



DATE:      Wednesday, November 22, 1989
TIME:      3:30 p.m.
LOCATION:  William G. Davis Computer Research Centre, Room 1302


ABSTRACT

A  major  goal  of   research   on   unsupervised   learning
procedures   is  to   discover  an  objective  function that
defines the quality of an  internal  representation  without
any  externally  supplied   information  about  the  desired
outputs of the system.  If such a function could  be  found,
it  should  allow  a  hierarchy  of  representations  to  be
organized  bottom-up in a time roughly linear in  the  depth
of the network.  This would allow much faster learning  than
supervised procedures  which  are  generally  very  slow  in
networks with many layers of hidden units. We propose that a
good  objective  for  perceptual   learning  is  to  extract
higher-order  features  that  are  coherent  across  time or
space.  This can  be   done   by   maximizing  the  explicit
mutual   information   between   parameters  extracted  from
spatially or temporally adjacent   parts   of   the   input.
Recent  results  obtained  by Sue Becker show that this kind
of objective function can  be  used  to  discover  depth  in
random-dot  stereograms.   The   approach  can be applied to
many other types of unsupervised  perceptual  learning.   In
particular,  it  should  be able to discover the  underlying
three-dimensional  shapes  of  objects  when presented  with
an ensemble of two-dimensional images.


Everyone is welcome.    Refreshments served.

rmvale@watcgl.waterloo.edu (Ruth Vale) (01/08/90)

ICR Colloquium

Practical Applications of Interior Point Algorithms

Dr. Anthony Vannelli
Department of Electrical & Computer Engineering
University of Waterloo

Wednesday, January 10, 1990 at 3:30 p.m.

Davis Centre 1302

Abstract

Since the introduction of Narendra Karmarkar's polynomial time
algorithm for solving linear programming problems in 1984, research 
in the mathematical optimization community has developed interior
point variants to solve quadratic programming and combinatorial
optimization problems.  In this talk we outline our own research
efforts to use an interior point algorithm to solve engineering-
related optimization problems that arise in such diverse areas as
water resource management, oil refinery multi-period planning
problems and VLSI circuit layout problems.  Our research on using
a dual affine scaling interior point algorithm has led to promising
resultsi in the outlined engineering areas.

A developed and flexible algorithm is described for solving these
large scale optimization problems.  In particular the effective management 
of the key projection step which is the bottleneck step in any interior
point code is described.  We indicate how to exploit the structure
of the underlying engineering design problem to minimize the
difficulties caused by the projection step.  Numerical results
are described which show that our interior point algorithm is 5-20 times
faster than the SIMPLEX code (MINOS) for solving these problems. 
Moreover, the algorithm becomes faster than the Simplex algorithm as
the problem size increases.

Everyone is welcome.  Refreshments served.